An intent-through physical execution method and system based on a dynamic semantic association base

By using the intent-to-physical execution method of the dynamic semantic association base, the architectural gap and performance bottleneck between the application layer and the underlying data storage layer of the large language model are resolved. This achieves physical-level isolation and decoupling of semantic topology and entity attribute data, improves system execution performance and security, reduces evolution costs, and ensures high system availability.

CN122346360APending Publication Date: 2026-07-07王平友
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
王平友
Filing Date
2026-04-13
Publication Date
2026-07-07

AI Technical Summary

Technical Problem

In existing technologies, there are architectural gaps and performance bottlenecks between the application layer and the underlying data storage layer of large language models, as well as pipeline fragmentation between the interaction layer and the scheduling layer, coupling between the staticization and storage of semantic structures, performance bottlenecks based on text intermediate states during the semantic-to-execution process, and gaps in security mechanisms.

Method used

The method adopts an intent-to-physical execution approach based on a dynamic semantic association base. It receives input instructions, performs isomorphic parsing to generate standardized semantic intents, verifies resource access permissions, uses the dynamic semantic association base for mapping and matching, generates a pure semantic structured expression, directly constructs a logical operator tree, and performs underlying resource scheduling to achieve physical-level isolation and decoupling between semantic topology and entity attribute data.

Benefits of technology

It achieves physical-level isolation and decoupling of semantic topology and entity attribute data, eliminates text parsing overhead and invalid I/O, unifies scheduling and improves system security, balances evolution cost and system high availability, and realizes dynamic decision-making of semantic operators and heterogeneous source normalization closed loop across the entire link.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure FT_1
    Figure FT_1
  • Figure FT_2
    Figure FT_2
  • Figure FT_3
    Figure FT_3
Patent Text Reader

Abstract

The application discloses an intention straight-through physical execution method and system based on a dynamic semantic association base, and belongs to the technical field of computer underlying execution engines and resource scheduling. In view of the problem that the dependence of existing large language model applications and underlying data storage on declarative text analysis leads to performance bottlenecks and strong coupling of semantic data, the application homogenizes heterogeneous input instructions into standard intentions, and maps the standard intentions in an independent semantic association base based on dynamic maintenance of a large model, storage of only entity topology and a physical mapping pointer. The system discards declarative text instructions, directly constructs a logical operator tree in memory based on a structured object serialization format output by a large model, and rewrites the logical operator tree into a physical execution operator tree in combination with semantic topology and underlying physical characteristics, and finally completes operations according to a mapping pointer and a bottom layer data storage layer. The application realizes physical-level isolation and decoupling of semantic topology and entity data, eliminates text analysis overhead, and significantly improves task scheduling efficiency and system security.
Need to check novelty before this filing date? Find Prior Art